minhh-nguyen opened a new pull request, #26105:
URL: https://github.com/apache/datafusion/pull/26105
## Which issue does this PR close?
<!--
We generally require a GitHub issue to be filed for all bug fixes and
enhancements and this helps us generate change logs for our releases. You can
link an issue to this PR using the GitHub syntax. For example `Closes #123`
indicates that this PR will close issue #123.
-->
- Closes #26071
## Rationale for this change
<!--
Why are you proposing this change? If this is already explained clearly in
the issue then this section is not needed.
Explaining clearly why changes are proposed helps reviewers understand your
changes and offer better suggestions for fixes.
Please explain the problem you are trying to solve in terms of the
user-visible
behavior, rather than the implementation.
For example, "The code in `foo.rs` doesn't handle nulls" is a symptom of the
implementation. "COUNT(DISTINCT) returns wrong results when the column
contains
nulls" is the user-visible problem.
-->
`EXPLAIN ANALYZE` reports `output_bytes` for every operator, and recording
that metric was costing more than the metric is worth. `BaselineMetrics`
called `get_record_batch_memory_size` on each output batch, which walks
every column and every child of a nested column and inserts each buffer's
address into a hash set so shared buffers are counted once.
That exact accounting is required for memory reservations, but it is not
required for a metric that is only displayed, and its cost grows with the
width and nesting of the schema rather than with the data. In a sampled
profile of a wide plan run through Sail on DataFusion 55.1,
`record_poll` -> `get_record_batch_memory_size` was about 12% of plan
execution.
This PR computes the metric without materializing `ArrayData`.
## What changes are included in this PR?
<!--
There is no need to duplicate the description in the issue here, but it is
sometimes worth providing a summary of the individual changes in this PR.
-->
- `BaselineMetrics` records `output_bytes` as the sum of each output
column's `Array::get_array_memory_size` instead of calling
`get_record_batch_memory_size`.
- `get_record_batch_memory_size` and `RecordBatchMemoryCounter` are
unchanged, so memory reservations keep exact, deduplicated accounting.
Every operator that reserves against batch memory still uses them directly.
- Document that `output_bytes` is now an approximation, and update the
metric description in `docs/source/user-guide/metrics.md`.
## What is the testing strategy for this PR?
<!--
We typically require tests for all PRs in order to:
1. Prevent the code from being accidentally broken by subsequent changes
2. Serve as another way to document the expected behavior of the code
Briefly describe how this PR is tested, and point to the specific tests you
added. For example: 'This new feature is covered by the `sqllogictest` cases
added in `foo.slt`'.
If this PR does not add tests, explain why. For example, if the change is
already covered by existing tests, please mention it.
You should also check the `codecov` bot reply on this PR to confirm the
changed code is exercised.
-->
no new test; output_bytes values are unchanged in existing sqllogictests;
note that hardcoded timings in explain_analyze.slt fail identically on main
## Are there any user-facing changes?
<!--
If there are user-facing changes then we may require documentation to be
updated before approving the PR.
If there are any breaking changes to public APIs, please add the `api
change` label.
-->
`EXPLAIN ANALYZE` now reports `output_bytes` as an approximation rather
than a buffer-deduplicated total. A buffer shared between columns or
between an array and its children is counted once per reference, and each
array's own in-memory struct is included. The metric description in
`docs/source/user-guide/metrics.md` was updated to say so. No query
results, public APIs, or memory accounting behavior change.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
